Date of Award

7-2015

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Computer Science

First Advisor

Raj Dasgupta

Second Advisor

Victor Winter

Third Advisor

Vyacheslav Rykov

Abstract

In this thesis, we consider the problem of efficient navigation by robots in initially unknown environments while performing tasks at certain locations. In initially unknown environments, the path plans might change dynamically as the robot discovers obstacles along its route. Because robots have limited energy, adaptations to the task schedule of the robot in conjunction with updates to its path plan are required so that the robot can perform its tasks while reducing time and energy expended. However, most existing techniques consider robot path planning and task planning as separate problems. This thesis plans to bridge this gap by developing a unified approach for navigating multiple robots in uncertain environments. We first formalize this as a problem called task ordering with path uncertainty (TOP-U) where robots are provided with a set of task locations to visit in a bounded environment, but the length of the path between a pair of task locations is initially known only coarsely by the robots. The robots must find the order of tasks that reduces the path length to visit the task locations. We then propose an abstraction called a task reachability graph (TRG) that integrates the robots task ordering and path planning. The TRG is updated dynamically based on inter-task path costs calculated by the path planner. A Hidden Markov Model-based technique calculates the belief in the current path costs based on the environment perceived by the robot’s sensors. We then describe a Markov Decision Process-based algorithm used by each robot in a distributed manner to reason about the path lengths between tasks and select the paths that reduce the overall path length to visit the task locations. We have evaluated our algorithm in simulated and hardware robots. Our results show that the TRG-based approach performs up to 60% better in planning and locomotion times with 44% fewer replans, while traveling almost-similar distances as compared to a greedy, nearest task-first selection algorithm.

Comments

A Thesis Presented to the Department of Computer Science And the Faculty of the Graduate College University of Nebraska In Partial Fulfillment Of the Requirements of the Degree Master of Science University of Nebraska at Omaha. Copyright 2015 Brad Woosley.

COinS